The quoted summaries of each of the three systems are confusing and I don’t feel like I have an understanding of them, except insofar as the word “Pavlovian” gives a hint. Can you translate more clearly, please?
Could you put it before the hard-to-parse explanations? It was nice to confirm my understanding, but it would have saved me a minute or two of effort if you’d put those first.
Sorry, I did not intend my comment to rub you the wrong way (or any of my previous comments that might have). FWIW, I think that you are doing a lot of good stuff for the SIAI, probably most of it invisible to an ordinary forum regular. I realize that you cannot afford spending extra two hours per post on polishing the message. Hopefully one of the many skills of your soon-to-be-hired executive assistant will be that of “optimizing presentation”.
It might be an interesting exercise to record predictions in a hidden-but-reliable form about karma of posts six months out, by way of calibrating one’s sense of how well-received those posts will be to their target community.
It’s still better than the posts I write in 2 hours! Did that 2 hours include the time spent researching, or were you just citing sources you’d already read for other reasons? In either case...not bad.
The first one incorporates information about past experiences into simplified models of the world, and then uses the models to steer decisions through search-space based upon a sort of back-of-the-envelope, hazy calculation of expected value. It’s a utility function, basically, as implemented by brain.
The second one also incorporates information about past experiences, but rather than constructing the dataset into a model and performing searches over it, it derives expectations directly from what’s remembered, and is insensitive to things like probability or shifting subjecting values.
The third one is sort of like the first in its basic operations (incorporate information, analyze it, make models) -- but instead of calculating expected values, it aims to satisfy various inbuilt “drives”, and sorts paths through search space based upon approach/avoid criteria linked to those drives.
The quoted summaries of each of the three systems are confusing and I don’t feel like I have an understanding of them, except insofar as the word “Pavlovian” gives a hint. Can you translate more clearly, please?
Or, to put it more simply:
Figure out what’s going on, and what actions maximize returns, and do them.
Do the thingy that worked before again!
Avoid the unpleasant thing and go to the pleasant thing. Repeat as necessary.
Added to the original post, credit given.
Could you put it before the hard-to-parse explanations? It was nice to confirm my understanding, but it would have saved me a minute or two of effort if you’d put those first.
Maybe give Luke a lesson or two on C^3 (clear, concise and catchy) summaries.
Note that I wrote this post in two hours flat and made little attempt to optimize presentation in this case.
Sorry, I did not intend my comment to rub you the wrong way (or any of my previous comments that might have). FWIW, I think that you are doing a lot of good stuff for the SIAI, probably most of it invisible to an ordinary forum regular. I realize that you cannot afford spending extra two hours per post on polishing the message. Hopefully one of the many skills of your soon-to-be-hired executive assistant will be that of “optimizing presentation”.
No worries!
Indeed. Much invisible work is required before optimization can occur. Invisible forging of skills precedes their demonstration.
For my own reference, here are the posts I tried to write well:
Secure Your Beliefs
Optimal Philanthropy for Human Beings
A Rationalist’s Tale
Existential Risk
Rationality Lessons Learned from Irrational Adventures in Romance
What Curiosity Looks Like
Can the Chain Still Hold You?
It might be an interesting exercise to record predictions in a hidden-but-reliable form about karma of posts six months out, by way of calibrating one’s sense of how well-received those posts will be to their target community.
It’s still better than the posts I write in 2 hours! Did that 2 hours include the time spent researching, or were you just citing sources you’d already read for other reasons? In either case...not bad.
Is 2 operant/Skinnerian conditioning, and 3 classical/Pavolvian conditioning?
If by “is” you mean “Do these correspond the underlying cognitive antecedents used in...”, then my answer is “it would seem so.”
The first one incorporates information about past experiences into simplified models of the world, and then uses the models to steer decisions through search-space based upon a sort of back-of-the-envelope, hazy calculation of expected value. It’s a utility function, basically, as implemented by brain.
The second one also incorporates information about past experiences, but rather than constructing the dataset into a model and performing searches over it, it derives expectations directly from what’s remembered, and is insensitive to things like probability or shifting subjecting values.
The third one is sort of like the first in its basic operations (incorporate information, analyze it, make models) -- but instead of calculating expected values, it aims to satisfy various inbuilt “drives”, and sorts paths through search space based upon approach/avoid criteria linked to those drives.
I like Jandila’s explanations.